Cancer-Associated Fibroblast-Derived CXCL11 Modulates Hepatocellular

Liu et al. Cell Death and Disease (2021) 12:260 https://doi.org/10.1038/s41419-021-03545-7 Cell Death & Disease

ARTICLE Open Access Cancer-associated ﬁbroblast-derived CXCL11 modulates hepatocellular carcinoma cell migration and tumor metastasis through the circUBAP2/ miR-4756/IFIT1/3 axis Gao Liu 1,JianSun1,Zhang-FuYang1, Cheng Zhou1, Pei-Yun Zhou1,Ruo-YuGuan1,Bao-YeSun1, Zhu-Tao Wang1, Jian Zhou1,JiaFan 1,Shuang-JianQiu1 and Yong Yi1

Abstract Cancer-associated fibroblasts (CAFs) are commonly acquired activated extracellular matrix (ECM)-producing myofibroblasts, a phenotypes with multiple roles in hepatic fibrogenesis and carcinogenesis via crosstalk with cohabitating stromal/cancer cells. Here, we discovered a mechanism whereby CAF-derived cytokines enhance hepatocellular carcinoma (HCC) progression and metastasis by activating the circRNA-miRNA-mRNA axis in tumor cells. CAFs secreted significantly higher levels of CXCL11 than normal fibroblasts (NFs), and CXCL11 also had comparatively higher expressions in HCC tissues, particularly in metastatic tissues, than para-carcinoma tissues. Both CAF-derived and experimentally introduced CXCL11 promoted HCC cell migration. Likewise, CAFs promoted tumor migration in orthotopic models, as shown by an increased number of tumor nodules, whereas CXCL11 silencing triggered a decrease of it. CXCL11 stimulation upregulated circUBAP2 expression, which was significantly higher in

1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; HCC tissues than para-carcinoma tissues. Silencing circUBAP2 reversed the effects of CXCL11 on the expression of IL- 1β/IL-17 and HCC cell migration. Further downstream, the IFIT1 and IFIT3 levels were signiﬁcantly upregulated in HCC cells upon CXCL11 stimulation, but downregulated upon circUBAP2 silencing. IFIT1 or IFIT3 silencing reduced the expression of IL-17 and IL-1β, and attenuated the migration capability of HCC cells. Herein, circUBAP2 counteracted miR-4756-mediated inhibition on IFIT1/3 via sponging miR-4756. miR-4756 inhibition reversed the effects induced by circUBAP2 silencing on the IL-17 and IL-1β levels and HCC cell migration. In orthotopic models, miR-4756 inhibition also reversed the effects on metastatic progression induced by silencing circUBAP2.

Introduction 80–90% of HCC develops with cirrhosis or fibrosis as a The occurrence of intrahepatic or systemic metastasis is required intermediate4, and approximately one-third of the main cause that result in poor prognosis of patients cirrhosis patients have a lifetime risk of developing HCC5. with advanced HCC1. Lung metastasis is the most com- Further elucidation on the mechanisms by which myofi- mon outcome of progression and one of the leading broblasts integrate with their cohabitating cancer cells causes of cancer deaths caused by HCC2,3. Approximately and how they bridge fibrosis and HCC through microenvironment remodeling might lead to the development of anticancer strategies. Correspondence: Shuang-Jian Qiu ([email protected])or Reportedly, cancer-associated fibroblasts (CAFs) origi- Yong Yi ([email protected]) nate from activated hepatic stellate cells and progressively 1 Department of Liver Surgery and Transplantation, Liver Cancer Institute and evolve into a major source of ECM as a consequence of Biomedical Research Center, Zhongshan Hospital, Fudan University, Shanghai 6–10 200032, PR China their accumulation in the stroma of HCC . Increasing Edited by G. Blandino

© The Author(s) 2021 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a linktotheCreativeCommons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 2 of 19

data now show that CAFs exert protumorigenic effects in CAFs and NFs. We found that the expression of CXCL11 tumor proliferation, migration, invasion, and EMT was most dramatically upregulated in CAFs when com- (epithelial–mesenchymal transformation) through cross- pared with NFs (Fig. 1B). Additionally, CXCL11 expres- talk with cancer cells via products of their paracrine sion was also upregulated in liver cirrhosis or liver cancer repertoire11,12. Over the last decade, a diverse range of tissues, compared to normal liver tissues, within the CAFs-derived chemokines, such as CCL2, CCL26, IL6, GSE14323 and GSE6764 datasets (Fig. S1A, B). CXCL1, and CXCL8, has been reported to be mechan- The intracellular and secreted levels of CXCL11 in istically related to tumor progression13,14. In addition, CAFs and NFs were also examined. The CXCL11 protein other mediators including IL-1, CXC, and CC chemokines levels were significantly upregulated in CAFs compared are also involved15,16. These factors might be the media- with NFs (Fig. 1C). The secreted CXCL11 level was also tors of CAFs contributing to carcinogenesis. higher in CAF culture medium than that in NF culture Recently, mounting evidence has shown that circRNAs medium (Fig. 1D). Consistently, the protein level of play a mediating role in tumor biology through various CXCL11 was increased in the nonmetastatic HCC tissue mechanisms related to the modulatory interactions samples and the metastatic HCC tissue samples compared among circRNAs, miRNAs, and mRNA17,18. An increas- with the in para-carcinoma tissues, and was further ing number of studies have noted that circRNA act as increased in the metastatic HCC tissue samples compared competing endogenous RNA (ceRNA) and subsequently with the nonmetastatic HCC tissue samples (Fig. 1E). counteract miRNA-induced effects on downstream Besides, the CXCL11 mRNA levels were increased in the mRNAs by sponging miRNAs19. For example, circRNA nonmetastatic HCC tissue samples and metastatic HCC 001306 promoted HCC progression by upregulating tissue samples compared with the para-carcinoma tissue CK16 by sponging miR-584-5p20. In addition, miRNAs, samples and were greater in the metastatic HCC tissue such as miR-101 and miR-346, have been demonstrated to samples than in the nonmetastatic HCC tissue samples play crucial roles in CAF-directed HCC progression21,22. (Fig. 1F). With anti-CXCL11, similar results were However, it has not yet been fully determined how the observed in IHC staining, in which CXCL11 proteins were circRNA-miRNA-mRNA network functions under the significantly increased in nonmetastatic HCC tissue effects of CAFs during HCC progression. samples and metastatic HCC tissue samples, and greater Given the roles of CAF-derived chemokines and non- in the metastatic HCC tissues than the nonmetastatic coding RNAs in tumor progression and metastasis, we HCC tissues (Fig. 1G). hypothesized that tumor progression directed by CAF- derived chemokines engages certain circRNA-miRNA- CAFs-derived CXCL11 affects the HCC cell phenotype mRNA axes in tumor cells. Herein, we show that CAF- To validate the specific roles of CAF-derived CXCL11 in derived CXCL11 was the most upregulated candidate the malignant behaviors of HCC cell, we collected dif- among the 17 differentially expressed chemokine candi- ferent conditioned media (CMs) for the HCC cell lines dates, and that elevation of CXCL11 in tumor micro- (MHCC-97H and Huh-7): control medium (control), environment promoted hepatocellular carcinoma cell NFs-derived conditioned medium (NFs-CM), CAFs- migration and tumor metastasis through the circUBAP2/ derived conditioned medium (CAFs-CM), CAFs-CM miR-4756/IFIT1/3 axis both in vivo and in vitro. supplemented with PBS (CAFs-CM/PBS), CAFs-CM supplemented with 5 ng/ml CXCL11 (CAFs-CM/ Results CXCL11 (5 ng/ml)), CAFs-CM supplemented with 10 ng/ CXCL11 expression in tissue samples and cancer- ml CXCL11 (CAFs-CM/CXCL11 (10 ng/ml)), CAFs-CM associated fibroblasts (CAFs) supplemented with 1 μg/ml IgG (CAFs-CM/IgG), and CAFs and NFs were isolated from cancer tissues and CAFs-CM supplemented with 1 μg/ml CXCL11 neu- noncancerous normal tissues, respectively. As shown in tralizing antibody (CAFs-CM/anti-CXCL11). Fig. 1A, relative fluorescence intensity of green fluores- Then, MHCC-97H and Huh-7 cells were cultured in cence representing the cytoskeletal protein α-SMA was different media accordingly and examined for morpho- dramatically enhanced in CAFs compared with NFs. logical changes and cell migration. Under SEM using Previous studies reported, CAFs secreted higher levels of different magnifications, MHCC-97H cells in the CAFs- CCL2, CCL5, CCL7, CXCL1612, CCL823, CXCL1124, CM, CAFs-CM/PBS, CAFs-CM/CXCL11 (5 ng/ml), CXCL1225, CXCL1326, and CXCL14 than NFs27. Besides, CAFs-CM/CXCL11 (10 ng/ml), and CAFs-CM/IgG both GSE14323 and GSE6764 indicated that these che- groups changed from flat-shaped to spindle-shaped, and mokines were highly expressed in HCV-induced liver the pseudopodia increased (Fig. 2A), but the appearance cirrhosis or liver cancer tissues. Therefore, we examined of HCC cells in NFs-CM and CAFs-CM/anti- the mRNA expression of CCL2, CCL5, CCL7, CCL8, CXCL11 showed no significant changes compared to that CXCL11, CXCL12, CXCL13, CXCL14, and CXCL16 in of the control (Fig. 2A).

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 3 of 19

Fig. 1 Expression of CXCL11 in tissue samples and cancer-associated fibroblasts (CAFs). A CAFs and normal fibroblasts were isolated from cancer tissues and noncancerous normal tissues, respectively, and identified for α-SMA protein content and distribution by Immunofluorescent (IF) staining. The relative fluorescence intensity was shown in the right panel. N = 5. B The mRNA expression of CXCL13, CXCL12, CCL2, CXCL11, CXCL16, CCL8, CXCL26, CXCL5, CXCL14, CCL5, and CCL7 was examined in CAFs and NFs using real-time PCR. C The protein levels of CXCL11 were examined in CAFs and NFs using Immunoblotting. D The contents of CXCL11 in the culture medium were determined by ELISA. E The protein levels of CXCL11 were examined in 12 cases of para-carcinoma tissues, 12 cases of nonmetastatic HCC tissues, and 12 cases of metastatic HCC tissues using Immunoblotting. F The mRNA expression of CXCL11 was examined in 12 cases of para-carcinoma tissues, 12 cases of nonmetastatic HCC tissues, and 12 cases of metastatic HCC tissues using real-time PCR. G The histopathological features of para-carcinoma tissues, nonmetastatic HCC tissues, and metastatic HCC tissues were examined by H&E staining. The protein contents of CXCL11 in para-carcinoma tissues, nonmetastatic HCC tissues, and metastatic HCC tissues were examined by Immunohistochemical (IHC) staining. *P < 0.05, **P < 0.01, compared with NFs group or para-carcinoma tissue group; ##P < 0.01, compared with the nonmetastatic HCC tissue group.

Next, HCC cell proliferation and migration in the dif- of HCC cells was signiﬁcantly augmented in CAFs-CM, ferent medium was examined. As shown in Fig. 2B–E and CAFs-CM/PBS, CAFs-CM/CXCL11 (5 ng/ml), CAFs- Fig. S2A, B, compared to those of the control and NFs- CM/CXCL11 (10 ng/ml), and CAFs-CM/IgG groups CM groups, proliferation, DNA synthesis and migration (*P < 0.05, **P < 0.01, #P < 0.05, ##P < 0.01), but these

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 4 of 19

Fig. 2 CAFs-derived CXCL11 affecting hepatocellular carcinoma (HCC) cell phenotype. HCC cell lines, MHCC-97H, and Huh-7, were cultured in control medium (Control), NFs-derived conditioned medium (NFs-CM), CAFs-derived conditioned medium (CAFs-CM), CAFs-CM added with PBS (CAFs-CM/PBS), CAFs-CM added with 5 ng/ml CXCL11 (CAFs-CM/CXCL11 (5 ng/ml)), CAFs-CM added with 10 ng/ml CXCL11 (CAFs-CM/CXCL11 (10 ng/ml)), CAFs-CM added with IgG (CAFs-CM/IgG), and CAFs-CM added with anti-CXCL11 (CAFs-CM/anti-CXCL11). A The morphological changes of HCC cell line MHCC-97H were examined by scanning electron microscopy (SEM) with different magniﬁcations; red arrows indicated the pseudopodia of HCC cells. B HCC cells viability was determined by MTT assays. C HCC cell DNA synthesis ability was determined by EDU assay. D HCC cell migration was examined using Transwell assay. E HCC cell migration was examined by wound healing assay. F The protein levels of Vimentin and Twist was examined by Immunoblotting. *P < 0.05, **P < 0.01, compared with the control group; #P < 0.05, ##P < 0.01, compared with the NFs-CM group; $P < 0.05, $$P < 0.01, compared with the CAFs-CM/PBS group; &P < 0.05, &&P < 0.01, compared with the CAFs-CM/IgG group.

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 5 of 19

showed no significant changes in CAFs-CM/anti-CXCL11 Bioinformatic analysis of circRNAs related to the functions group. Compared with those of CAFs-CM and CAFs-CM/ of CAF-secreted CXCL11 in aggressive HCC PBS groups, HCC cell proliferation, DNA synthesis and We performed RNA sequencing on HCC cells with/ migration ability were significantly promoted in CAFs- without CXCL11 to validate the possibility that circRNAs CM/CXCL11 (5 ng/ml) and CAFs-CM/CXCL11 contribute to the oncogenic effects of CAF-secreted (10 ng/ml) groups ($P < 0.05, $$P < 0.01; Fig. 2B–E). CXCL11 in aggressive HCC. A total of 3124 circRNAs Compared with those of CAFs-CM/IgG group, HCC cell were detected and 24 were significantly differentially proliferation, DNA synthesis and migration were sig- expressed (|logFC| > 0.56, p < 0.05); 11 were down- nificantly inhibited in CAFs-CM/anti-CXCL11 group (&P regulated and 13 were upregulated (Fig. 3A, B). According < 0.05, &&P < 0.01; Fig. 2B–E). to GSE97332, a total of 1180 circRNAs exhibited differ- We then examined the changes in the Vimentin and ential expression within HCC tissue samples compared Twist protein levels in different groups in search of bio- with normal liver tissue samples; 641 were upregulated factors related to tumor migration. Compared to those of and 539 were downregulated (Fig. 3C). The top 10 the control and NFs-CM groups, Vimentin and Twist were upregulated and downregulated circRNAs were shown in dramatically increased in CAFs-CM, CAFs-CM/PBS, CAFs- Fig. 3D. The intersection of RNA sequencing data and CM/CXCL11 (5 ng/ml), CAFs-CM/CXCL11 (10 ng/ml), GSE97332 included circRERE and circUBAP2 (upregu- and CAFs-CM/IgG groups (*P <0.05,**P <0.01, #P <0.05, lated) and circPPP6R3 (downregulated), and according to ##P < 0.01; Fig. 2E and Fig. S2C), but Vimentin and Twist in all four datasets, circUBAP2 was significantly upregulated the CAFs-CM/anti-CXCL11 group showed no significant in the CXCL11-treated HCC cells or HCC tissues changes. And, compared with those of the CAFs-CM and (Fig. 3E). CAFs-CM/PBS groups, Vimentin and Twist were sig- Next, we treated the HCC cell lines MHCC-97H and nificantly increased in the CAFs-CM/CXCL11 (5 ng/ml) Huh-7 with CXCL11 (10 ng/ml) and performed real-time and CAFs-CM/CXCL11 (10 ng/ml) groups ($P <0.05,$$P < PCR to determine the circRERE, circUBAP2, and 0.01; Fig. 2E and Fig. S2C). Compared with those of the circPPP6R3 expression levels. Figure 3F shows that only CAFs-CM/IgG group, Vimentin and Twist were sig- circUBAP2 expression was dramatically induced by nificantly decreased in the CAFs-CM/anti-CXCL11 group CXCL11 stimulation. The expression of circUBAP2 was (&P <0.05, &&P < 0.01; Fig. 2E and Fig. S2C). significantly upregulated in the CXCL11-treated MHCC- 97H cells compared with the untreated MHCC-97H cells CAFs-derived CXCL11 affects the tumor growth in BALB/c (Fig. 3G) and increased in the HCC tissue samples com- nude mice with orthotopically implanted tumors pared with the noncancerous tissue samples according to To investigate the in vivo effects of CAFs-derived GSE94508, GSE97332, and GSE78520 (Fig. 3H). More- CXCL11, we first generated orthotopic implantation over, circUBAP2 expression was significantly increased in tumor models in mice. Mice were randomly assigned to the nonmetastatic HCC tissue samples and metastatic five groups: the Huh-7 cells group (n = 6), the Huh-7 HCC tissue samples compared with the para-carcinoma cells + NFs group (n = 6), the Huh-7 cells + CAFs tissue samples and was increased in the metastatic HCC (untransfected; n = 6) group, the Huh-7 cells + CAFs tissue samples than that within nonmetastatic HCC tissue (transfected with sh-NC; n = 6) group, and the Huh-7 samples (Fig. 3I). Thus, circUBAP2 was chosen for further cells + CAFs (transfected with sh-CXCL11; n = 6) group. experiments. Fourteen days after cell injection, the mice were then anesthetized and sacrificed, the livers were collected (Fig. circUBAP2 is involved in the effects of CXCL11 on HCC cells S3A), and the tumor nodules were counted. As shown in Since CXCL11 significantly induces circUBAP2 expres- Fig. S3B, there were more tumor nodules in the Huh-7 sion, we achieved the silencing of circUBAP2 in MHCC- cells + CAFs and Huh-7 cells + CAFs (sh-NC) groups 97H and Huh-7 cells by transfecting siRNA targeting cir- than that in the Huh-7 cells and Huh-7 cells + NFs cUBAP2 (si-circUBAP2-1/si-circUBAP2-2); si-NC was groups; there were fewer tumor nodules in the Huh-7 transfected as a negative control, as confirmed by real-time cells + CAFs group than that in the Huh-7 cells + CAFs PCR; and si-circUBAP2-1 was chosen for further experi- and Huh-7 cells + CAFs (sh-NC) groups. Livers from ments because it had better transfection efficiency than the different groups were evaluated by H&E staining for his- other siRNA (Fig. 4A). Then, we divided MHCC-97H and topathological characteristics (Fig. S3C). In orthotopic Huh-7 cells into five groups: the control group (untreated, tumor tissues, CXCL11 protein expression was dramati- untransfected), si-NC group (untreated, transfected with si- cally higher in the Huh-7 cells + CAFs and Huh-7 cells + NC), CXCL11 + si-NC group (treated with 10 ng/ml CAFs (sh-NC) groups compared with those in the Huh-7 CXCL11, transfected with si-NC), si-circUBAP2 group cells + CAFs and Huh-7 cells + CAFs (sh-NC) groups (untreated, transfected with si-circUBAP2), and CXCL11 + (Fig. S3D, E). si-circUBAP2 group (treated with 10 ng/ml CXCL11,

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 6 of 19

Fig. 3 Selection of circRNA related to the functions of CAF-secreted CXCL11 on HCC cell aggressiveness. A, B Volcano plot and hierarchical clustering heatmap showing differentially expressed circRNAs in CXCL11-treated MHCC-97H based on RNA sequencing. C, D Volcano plot and hierarchical clustering heatmap showing differentially expressed circRNAs between HCC and normal tissues based on GSE97332. E A schematic diagram showing the selection of circRNAs related to the functions of CAF-secreted CXCL11 on HCC cell aggressiveness based on our RNA sequencing data and online microarray expression proﬁles. These datasets intersected at the following circRNAs: circUBAP2 (upregulated), circRERE (upregulated), and circPPP6R3 (downregulated). F HCC cell lines MHCC-97H and Huh-7 were treated with CXCL11 (10 ng/ml) and examined for the expression of circUBAP2, circRERE, and circPPP6R3 by real-time PCR. G The expression of circUBAP2 in CXCL11-treated or untreated MHCC-97H and Huh-7 cells. H The expression of circUBAP2 in normal liver tissues and liver cancer tissues based on GSE97332, GSE94508 and GSE78520. I The expression of circUBAP2 was examined in 12 cases of nonmetastatic HCC tissues, metastatic HCC tissues, and para-carcinoma tissues by real-time PCR. **P < 0.01, ##P < 0.01.

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 7 of 19

Fig. 4 (See legend on next page.) transfected with si-circUBAP2). HCC cells were examined cells was signiﬁcantly promoted in the CXCL11 + si-NC for cell migration and related markers. Compared to that of group but inhibited in the si-circUBAP2 group; the pro- the control and si-NC groups, the migratory ability of HCC motive effects of CXCL11 stimulation on HCC cell

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 8 of 19

(see figure on previous page) Fig. 4 The function of circUBAP2 in the effects of CAFs-derived CXCL11 on HCC cells and the selection of mRNA related to the functions of circUBAP2. A Silencing of circUBAP2 was achieved in MHCC-97H and Huh-7 cells by transfecting small interfering RNA targeting circUBAP2 (si- circUBAP2-1 or si-circUBAP2-2). Si-NC was transfected as a negative control. The transfection efficiency was confirmed by real-time PCR and si- circUBAP2-1 was chosen for further experiments for its better transfection efficiency. Then, MHCC-97H and Huh-7 cells were transfected with si- circUBAP2 and examined for B Cell migration by Transwell assay; C Cell migration by wound healing assay; D The protein levels of Vimentin and Twist was examined by Immunoblotting; E The concentrations of IL-1β and IL-17 in the culture medium by ELISA. F, G The Volcano plot and hierarchical clustering heatmap showing differentially expressed mRNAs in CXCL11-treated MHCC-97H based on RNA sequencing. Upregulated genes were applied for H Kyoto Encyclopedia of Genes and Genomes (KEGG) signaling enrichment analysis and I Gene Ontology (GO) of biological process enrichment analysis. J The expression of IFIT1/IFIT3 in CXCL11-treated and untreated HCC cells based on RNA sequencing data. *P < 0.05, **P < 0.01, compared with the control group; #P < 0.05, ##P < 0.01, compared with the si-NC group; $P < 0.05, $$P < 0.01, compared with the CXCL11 + si- circUBAP2 group.

migration were significantly reversed by circUBAP2 HCC cells by CXCL11 stimulation (Fig. 4J). Thus, we silencing (Fig. 4B, C and Fig. S4A, B). For the cell migration- speculated that IFIT1/IFIT3 might participate in the related markers, compared with those in the control and functions of CAF-secreted CXCL11 and circUBAP2 in si-NC groups, the Vimentin and Twist protein levels were HCC cells. dramatically increased in the CXCL11 + si-NC group but decreased in the si-circUBAP2 group; the effects of Effects of IFIT1/IFIT3 on HCC cell phenotype CXCL11 stimulation on the Vimentin and Twist protein Firstly, we determined the expression of IFIT1 and levels were partially reversed by circUBAP2 silencing (Fig. IFIT3 in tissues. We found that the mRNA levels of IFIT1 4D and Fig. S4C). Moreover, the concentrations of the IL-17 and IFIT3 were dramatically increased in the nonmeta- signaling cytokines, IL-1β and IL-17, were examined. Simi- static HCC tissue samples and the metastatic HCC tissue larly, compared with those in the control and si-NC groups, samples compared with the para-carcinoma tissue sam- the medium concentrations of IL-1β and IL-17 were dra- ples, and increased in the metastatic HCC tissue samples matically higher in the CXCL11 + si-NC group but lower in compared with the nonmetastatic HCC tissue samples the si-circUBAP2 group; the effects of CXCL11 stimulation (Fig. 5A). The protein levels of IFIT1 and IFIT3 showed a on the IL-1β and IL-17 medium concentrations were similar expression trend as their mRNA levels in HCC partially reversed by circUBAP2 silencing (Fig. 4E). tissues (Fig. 5B). Figure 5C shows that circUBAP2 knockdown dramatically downregulated IFIT1 and IFIT3 Bioinformatic analysis of mRNAs related to the functions of mRNA expression in MHCC-97H and Huh-7 cells. CAF-secreted CXCL11 and circUBAP2 in aggressive HCC Next, we transfected small interfering RNA targeting RNA-Seq was performed on CXCL11-stimulated or IFIT1 or IFIT3 to achieve IFIT1 or IFIT3 silencing in unstimulated HCC cells to identify differentially expres- MHCC-97H and Huh-7 cells. As confirmed by real-time sed genes. According to our RNA sequencing data, a total PCR, si-IFIT3-1 and si-IFIT1-2 were chosen for further of 11622 mRNAs were detected and 211 mRNAs exhib- experiments due to their best transfection efficiency (Fig. ited significant differential expression; of them, 142 were 5D). As revealed by Transwell and wound healing assays, increased and 69 were decreased in HCC cells via IFIT1 silencing or IFIT3 silencing was sufficient to inhibit CXCL11 stimulation (Fig. 4F). The top 15 upregulated or HCC cell migration (Fig. 5E, F). Consistently, IF staining downregulated genes were shown in the clustered heat and Immunoblotting also showed that IFIT1 silencing map (Fig. 4G). These upregulated genes were subjected to or IFIT3 silencing was sufficient to decrease the expres- KEGG signaling enrichment analysis and Gene GO sions of Vimentin and Twist (Fig. 5G, H). Moreover, enrichment analysis. As shown in Fig. 4H, KEGG IFIT1 silencing or IFIT3 silencing significantly reduced enrichment annotation (https://www.genome.jp/kegg/) the medium concentrations of IL-17 and IL-1β suggested that these upregulated genes were significantly (Fig. 5I). These data indicated that IFIT1 silencing or enriched in IL-17 signaling, Nod-like receptor signaling, IFIT3 silencing attenuates the HCC cell aggressiveness. herpes simplex infection pathway and so on (Table S2). GO enrichment analysis of the biological process also miR-4756 directly binds to circUBAP2 and the 3′-UTR of suggested that these upregulated genes were obviously IFIT1/IFIT3 enriched signaling related to the cellular response to type Here, we continued to search for miRNAs that might I interferon (Fig. 4I). Among these differentially expressed mediate the crosstalk between circUBAP2 and IFIT1/IFIT3 genes, interferon-induced protein with tetratricopeptide with ENCORI and TargetScan. These two predicted sets of repeats 1 (IFIT1) and IFIT3 are interferon signaling miRNAs intersected at hsa-miR-4756-5p, hsa-miR-1321, related genes28 and were both significantly upregulated in hsa-miR-552-3p, hsa-miR-3611, and hsa-miR-24-3p

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 9 of 19

Fig. 5 Effects of IFIT1/IFIT3 on HCC cell phenotype. A The mRNA expression of IFIT1 and IFIT3 was examined in 12 cases of nonmetastatic HCC tissues, metastatic HCC tissues, and para-carcinoma tissues by real-time PCR. B The protein contents and distribution of IFIT1 and IFIT3 were examined in nonmetastatic HCC tissues, metastatic HCC tissues, and para-carcinoma tissues by IHC staining. C MHCC-97H and Huh-7 cells were transfected with si- circUBAP2 and examined for the mRNA expression of IFIT1 and IFIT3 by real-time PCR. D IFIT1 or IFIT3 silencing was achieved in MHCC-97H and Huh-7 cells by transfecting small interfering RNA targeting IFIT1 or IFIT3 (si-IFIT1-1 or si-IFIT1-2; si-IFIT3-1 or si-IFIT3-2). Si-NC was transfected as a negative control. The transfection efficiency was confirmed by real-time PCR and si-IFIT3-1 and si-IFIT1-2 were chosen for further experiments due to better transfection efficiency. Next, MHCC-97H and Huh-7 cells were transfected with si-NC, si-IFIT1, or si-IFIT3, and examined for E Cell migration by Transwell assay; F Cell migration by wound healing assay; G Cellular protein content and distribution of Vimentin was examined by IF staining; H The protein levels of Vimentin and Twist was examined by Immunoblotting; I The concentrations of IL-1β and IL-17 in the culture medium by ELISA. *P <0.05,**P < 0.01, compared with the si-NC group.

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 10 of 19

Fig. 6 (See legend on next page.)

(Fig. 6A). The expression of miR-4756-5p, miR-1321, miR- showninFig.6A, only miR-4756 was signiﬁcantly down- 552-3p, miR-3611, and miR-24-3p was examined in MHCC- regulated in both HCC cell lines upon CXCL11 stimulation. 97H and Huh-7 cells with/without CXCL11 stimulation. As Moreover, only miR-4756 was dramatically increased in

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 11 of 19

(see figure on previous page) Fig. 6 miR-4756 directly binds to circUBAP2 and the 3′-UTR of IFIT1/IFIT3. A ENCORI was used to predict miRNAs that circUBAP2 might target, and TargetScan was used to predict miRNAs that might target IFIT1/IFIT3. These two sets of miRNAs intersected at hsa-miR-4756-5p, hsa-miR-1321, hsa-miR-552-3p, hsa-miR-3611, and hsa-miR-24-3p. The expression of miR-4756-5p, miR-1321, miR-552-3p, miR-3611, and miR-24-3p was examined in MHCC-97H and Huh-7 cells with or without CXCL11 treatment. miR-4756 was chosen for further experiments due to its downregulation in both HCC cell lines in response to CXCL11 stimulation. B MHCC-97H and Huh-7 were transfected with si-circUBAP2 and examined for the expression of miR- 4756 by real-time PCR. C The expression of miR-4756 was examined in 12 cases of nonmetastatic HCC tissues, metastatic HCC tissues, and para- carcinoma tissues by real-time PCR. D miR-4756 overexpression or inhibition was achieved in MHCC-97H and Huh-7 cells by transfecting miR-4756 mimics or miR-4756 inhibitor; the transfection efficiency was confirmed by real-time PCR. E MHCC-97H and Huh-7 cells were transfected with miR- 4756 mimics or miR-4756 inhibitor and examined for the protein levels of IFIT1 and IFIT3 by Immunoblotting. F–H Wild- and mutant-type circUBAP2, IFIT1 3′-UTR, or IFIT3 3′-UTR luciferase reporter plasmids were constructed as described and named as wt-circUBAP2/mut-circUBAP2, wt-IFIT1/mut- IFIT1, and wt-IFIT3/mut-IFIT3. These plasmids were co-transfected in 293T cells with miR-4756 mimics or miR-4756 inhibitor and the luciferase activity was determined. *P < 0.05, **P < 0.01, compared to si-NC + inhibitor NC group. ##P < 0.01, compared to si-cirUBAP2+miR-4756 inhibitor group. both HCC cell lines in the circUBAP2-silenced MHCC-97H 4756 inhibitor and determined the IFIT1 and IFIT3 pro- and Huh-7 cells (Fig. 6B). And further experiments con- tein levels. As shown in Fig. 7A, circUBAP2 silencing firmed the expression of miR-4756 was significantly down- significantly decreased, whereas miR-4756 inhibition regulated in the nonmetastatic HCC tissue samples and increased, the IFIT1 and IFIT3 protein levels; miR-4756 metastatic HCC tissue samples compared with the para- inhibition significantly attenuated the effects of circU- carcinoma tissue samples and increased in the metastatic BAP2 knockdown. In summary, circUBAP2 acts as a HCC tissue samples than that in the nonmetastatic HCC ceRNA for miR-4756, competing with IFIT1/IFIT3 for tissue samples (Fig. 6C). miR-4756 binding and counteracting miR-4756-mediated Next, miR-4756 overexpression or inhibition was IFIT1/IFIT3 inhibition. achieved in MHCC-97H and Huh-7 cells by transfecting Next, the coeffects of circUBAP2 and miR-4756 on miR-4756 mimics or miR-4756 inhibitor, as confirmed by HCC cell aggressiveness were examined. As shown in real-time PCR (Fig. 6D). In MHCC-97H and Huh-7 cells, Fig. 7B, the medium concentrations of IL-1β and IL-17 miR-4756 overexpression decreased, while miR-4756 were significantly reduced by circUBAP2 silencing inhibition elevated the IFIT1 and IFIT3 protein levels but increased by miR-4756 inhibition; the effects of (Fig. 6E). Thus, circUBAP2 negatively modulates miR- circUBAP2 silencing were partially reversed by miR-4756 4756 expression, and miR-4756 negatively modulates inhibition. Transwell and wound healing assays indicated IFIT1 and IFIT3 expression. that circUBAP2 silencing inhibited HCC cell migration, Then, we validated the binding of miR-4756 to circU- whereas miR-4756 inhibition promoted HCC cell migra- BAP2 and IFIT1/IFIT3 with dual-luciferase reporter tion; miR-4756 inhibition significantly attenuated the assay. We constructed wild-type and mutant-type circU- effects of circUBAP2 silencing (Fig. 7C, D). Consistently, BAP2 and IFIT1 3′-UTR and IFIT3 3′-UTR luciferase circUBAP2 silencing decreased, while miR-4756 inhibi- reporter plasmids as described and named them wt-cir- tion elevated Vimentin and Twist levels; miR-4756 inhi- cUBAP2/mut-circUBAP2, wt-IFIT1/mut-IFIT1, and wt- bition significantly attenuated the effects of circUBAP2 IFIT3/mut-IFIT3 (Fig. 6F–H). We co-transfected these silencing (Fig. 7E, F). In summary, circUBAP2 participates plasmids into 293T cells with miR-4756 mimics/inhibitor, in the functions of CAFs-derived CXCL11 through the and examined luciferase activity. Figure 6F–H shows that circUBAP2/miR-4756/IFIT1/IFIT3 axis. miR-4756 overexpression dramatically decreased, while miR-4756 inhibition increased, the luciferase activity of Combined effects of circUBAP2 and miR-4756 on tumor wt-circUBAP2, wt-IFIT1, and wt-IFIT3; after we mutated growth and lung metastasis in the BALB/c mice with the putative miR-4756 binding site, miR-4756 over- orthotopically implanted tumors expression or miR-4756 inhibition failed to change the Finally, we established an orthotopic implantation luciferase activity of mut-circUBAP2, mut-IFIT1, or tumor model by injecting Huh-7 cells coinfected with sh- mut-IFIT3 (Fig. 6F–H). Thus, miR-4756 is a direct target NC/sh-circUBAP2 and anti-NC/anti-miR-4756 lentivirus of circUBAP2 and IFIT1/IFIT3 are direct targets of into the left liver lobe of mice to investigate their effects miR-4756. on lung metastasis. Four weeks after the injection, anesthetized mice were sacrificed, and livers and lungs were Combined effects of circUBAP2 and miR-4756 on IFIT1/ collected for further experiments (Fig. 8A, C). As shown IFIT3 expression and the HCC cell phenotype in Fig. 8B, the mice in the sh-circUBAP2 + anti-NC group To confirm whether circUBAP2, miR-4756, and IFIT1/ developed the fewest liver tumor nodules, whereas the IFIT3 could form a regulatory axis, we co-transfected mice in the sh-NC + anti-miR-4756 group developed the MHCC-97H and Huh-7 cells with si-circUBAP2 and miR- most liver tumor nodules; the liver tumor nodules

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 12 of 19

Fig. 7 Combined effects—effects of circUBAP2 and miR-4756 on IFIT1/IFIT3 expression and HCC cell phenotype. MHCC-97H and Huh-7 cells were co-transfected with si-circUBAP2 and miR-4756 inhibitor and examined for A The protein levels of IFIT1 and IFIT3 by Immunoblotting; B The concentrations of IL-1β, and IL-17 in the culture medium by ELISA; C Cell migration by Transwell assay; D Cell migration by wound healing assay; E The cellular protein content and distribution of Vimentin was examined by IF staining; F The protein levels of Vimentin and Twist was examined by Immunoblotting. *P < 0.05, **P < 0.01, compared with the si-NC + inhibitor NC group; ##P < 0.01, compared with the si-circUBAP2 + miR-4756 inhibitor group.

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 13 of 19

number in the mice from the sh-circUBAP2 + anti-miR- confirmed a key role of CXCL11 in mediating the 4756 group was almost the same as that in the sh-NC + crosstalk between CAFs and HCC progression. anti-NC group. Similar results were observed for lung Increasing evidence demonstrates the possible involve- metastatic nodules (Fig. 8D), as confirmed by H&E ment of noncoding RNAs in the actions of cytokines staining, in which circUBAP2 silencing attenuated, secreted by CAFs. However, previous studies mostly whereas miR-4756 inhibition enhanced, the lung meta- focused on the mechanisms underlying tumor progres- static ability (Fig. 8E). Additionally, as shown by IHC sion by long noncoding RNAs and miRNAs19, while less is staining and immunoblotting, the protein levels of IFIT1/ known about the functions and mechanisms of CAF- 3 were decreased by circUBAP2 knockdown but elevated related circRNAs in HCC. Herein, we performed RNA by miR-4756 inhibition; miR-4756 inhibition significantly sequencing and analyzed the differentially expressed cir- attenuated the effects of circUBAP2 knockdown on cRNAs and mRNAs between CXCL11-stimulated and IFIT1/3 levels in the orthotopically implanted tumors unstimulated HCC cells. Of these differentially expressed (Fig. 8F, G). In mouse liver tumors, sh-circUBAP2 circRNAs, circUBAP2 showed a much greater potential to downregulated circUBAP2 expression and upregulated facilitate HCC progression than other circRNAs accord- miR-4756 expression, whereas the miR-4756 inhibitor ing to its expression status in HCC and its roles in pre- exerted the opposite effects; the effects of sh-circUBAP2 vious studies of other diseases from the literature. The were partially reversed by miR-4756 inhibitor (Fig. 8F). oncogenic role of circUBAP2 has already been reported Consistently, sh-circUBAP2 decreased, whereas the miR- and found to work as a ceRNA. circUBAP2 modulates 4756 inhibitor elevated Vimentin and Twist in the mouse pancreatic adenocarcinoma tumorigenesis by regulating liver tumors; the effects of sh-circUBAP2 were partially the levels of hsa-miR-494 and the expression of CXCR4, reversed by the miR-4756 inhibitor (Fig. 8G). HIF1A, ZEB1, SDC1, and TWIST132. In addition, circU- BAP2 modulates cancer cell malignancy through the Discussion miR-641/YAP1 axis and miR-361-3p/SOX4 axis in CAFs overexpress proinflammatory genes and mod- osteosarcoma and cervical cancer, respectively33,34. Con- ulate the inflammatory microenvironment in the fibrotic sistently, our results showed that circUBAP2 silencing was liver, which could lead to tumor development and sufficient to attenuate HCC cell migration, as well as the facilitate tumor progression10,12. Previously, Fang et al.29 promotive effects of CXCL11 on HCC cell migration. demonstrated that activated CAFs further promote Intriguingly, our RNA sequencing also identified a cancer progression by secreting proinflammatory cyto- number of mRNAs deregulated in the CXCL11- kines, including IL-6 and IL-8. Herein, we detected stimulated HCC cells, and the signaling and function dramatic increases in CXCL11 levels by CAFs when annotation results suggested that upregulated mRNAs comparing the expression levels of different cytokines were dramatically enriched in IL-17 signaling and the cell secreted by CAFs and NFs. CXCL11 is part of the response to interferon. These annotations were consistent superfamily of chemotactic cytokines and contributes to with the fact that CXCL11 belongs to the class of IFN-γ inflammatory responses, often through selective T cell inducible chemokines, which are notably elevated in the recruitment to inflammatory sites. The results from process of hepatic inflammation35. Likewise, Helbig other studies demonstrated a significantly elevated level et al.36 revealed that stimulation of Huh-7 cells with IFN-γ of CXCL11 in HCC30,31 and indicated a role of CXCL11 could also produce high levels of CXCL11 in vitro, indi- in regulating the self-renewal and tumorigenicity of liver cating that IFN-γ might serve as a key regulatory factor of tumor-initiating cells31. Consistently, we also observed CXCL11 in malignant and inflammatory hepatocytes. that in vitro cultivation with conditioned media of CAFs Given these findings, we next focused on the interferon remarkably promoted the migratory capacity of HCC signaling related gene IFIT1/337 for further investigations. cells. And when additional CXCL11 was introduced into si-circUBAP2 downregulated IFIT1/3 expression, further conditioned media, cancer cell migration was further confirming the role of circUBAP2 in regulating the enhanced. These in vitro findings were then confirmed expression of IFIT1/3 as a ceRNA. Indeed, silencing either by in vivo investigations that CAFs increased the num- IFIT1 or IFIT3 reduced the IL-17 and IL-1β levels and ber of tumor metastases in the mice bearing Huh-7 cell- attenuated the aggressiveness of HCC cells. derived tumors; when CXCL11 was silenced, the effects We then intended to identify miRNAs that might form a of CAFs on tumor progression were significantly atte- ceRNA regulatory network with circUBAP2 and IFIT1/3, nuated. The present findings were consistent with pre- and with predictive tools, we succeeded in identifying vious findings by Liu et al.12 that CAFs promote the miR-4756 as a potential participant. By competing with migration and invasion of HCC cells in vitro and facil- IFIT1/3 for miR-4756 binding, circUBAP2 attenuated itate HCC metastasis to the bone, brain and lung in miR-4756-induced IFIT1/3 inhibition. Despite a previous NOD/SCID mice. These results fundamentally report that exosomal miR-320a38 and miR-2139 from

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 14 of 19

Fig. 8 Combined effects of circUBAP2 and miR-4756 on tumor growth and lung metastasis in the orthotopically implanted tumor model in BALBc nude mice. A, C The orthotopic implantation tumor model was conducted by injecting Huh-7 cells coinfected with sh-NC/sh-circUBAP2 and anti-NC/anti-miR-4756 lentivirus into the left liver lobe of mice. After 5 weeks, anesthetized mice were sacriﬁced and livers and lungs were collected for further experiments. B Tumor numbers in mice livers from different groups were counted. D Lung metastatic nodules in mice from different groups were counted. E The histopathological characteristics of mice livers and lungs were evaluated by H&E staining. The protein content and distribution of IFIT1/3 in mice liver tumors were examined by IHC staining. F The expression levels of circUBAP2 and miR-4756 in mice livers from different groups were examined by qRT-PCR. G The protein levels of IFIT1/3, Vimentin, and Twist in mice livers tumors from different groups were examined by Immunoblotting. H A schematic graph of the mechanism. CXCL11 acts as a key mediator integrating CAFs and CAFs-cohabitating HCC cells, and also work as an extracellular remodeler promoting HCC migration and metastasis through the activation of circUBAP2/miR-4756/IFIT1/3 in tumor cells. **P < 0.01, compared with the sh-NC + anti-NC group; ##P < 0.01, compared with the sh-circUBAP2 + anti-miR-4756 group.

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 15 of 19

CAFs mediate the crosstalk between CAFs and HCC cells, CD31, AFP and pan-cytokeratin, respectively. Fibroblasts we are the first to demonstrate that CXCL11 stimulation isolated from tumor tissues were defined as CAFs, and induces the deregulation of miR-4756 in HCC cells, which those isolated from nontumor counterparts were defined might subsequently be involved in the role of CXCL11 in as NFs. CAFs and NFs of less than ten generations were HCC cell migration. In contrast to circUBAP2 and IFIT1/ used for experiments. 3, miR-4756 expression was significantly downregulated in HCC cells in response to CXCL11 stimulation. Preparation of conditioned medium (CM) Regarding cellular functions, miR-4756 inhibition pro- CAFs or NFs were seeded on six-well plates at a den- moted HCC cell migration, and the effects of sity of 1 × 105 cells/well. After 24 h of cell seeding, the circUBAP2 silencing were partially reversed by miR-4756 medium was removed, the cells were washed once with inhibition. miR-4756 inhibition also reversed the effects of PBS, and 1 ml of serum-free medium was added to each circUBAP2 silencing on metastatic progression in mice well. Then, the cells were incubated at 37 °C for 24 h, the bearing Huh-7 cell-derived tumors. CM was collected, and the collected CM was passed In summary, our work suggests a mechanism whereby through a 0.2 μm membrane syringe filter to remove all CXCL11 acts as a key mediator integrating CAFs and cells and cell debris. Then, the CM was collected for CAF-cohabitating cancer cells, and functions as an further use. extracellular remodeler promoting HCC migration and metastasis through the activation of circUBAP2/miR- Immunofluorescence (IF) staining 4756/IFIT1/3 in tumor cells (Fig. 8H). HCC cells and collected CAFs and NFs were fixed with 4% paraformaldehyde for 20 min, washed with Materials and methods PBS and permeabilized with 0.1% Triton X-100 for Clinical sampling 10 min. Then, the cells were blocked in PBS containing Twelve HCC samples with metastasis (metastatic HCC; 10% BSA at room temperature for 1 h. At the end of the including portal vein invasion, extrahepatic distant blocking, the CAFs and NFs were incubated with anti- metastasis, capsule invasion, vascular tumor thrombus, α-SMA (55135-1-AP; Proteintech, Wuhan, China) or portal vein tumor thrombus, lymph node metastasis) and anti-Vimentin (10366-1-AP, Proteintech) at 4 °C 12 HCC samples without metastasis (nonmetastatic HCC) overnight. HCC cells were incubated with anti- were collected. Postoperative pathology confirmed the Vimentin (10366-1-AP, Proteintech) at 4 °C over- diagnosis of HCC, and none of the patients received night. Then, the cells were stained with an appropriate preoperative radiotherapy or TACE (transarterial che- FITC-conjugated secondary antibody (1:1000) in a dark moembolization). In addition, for 12 cases, adjacent room at 37 °C for 1 h. Then, DAPI was used to stain the hepatic tissues were collected from these HCC patients nucleus, and the cover glass was sealed with Antifade with metastases as negative controls. All samples were Mounting Medium. collected from patients who received surgical resection at Zhongshan Hospital, Fudan University. Informed consent HCC cell line was signed and obtained from all patients enrolled. All A metastatic HCC cell line, MHCC-97H, was estab- clinical sampling was performed with the approval of the lished and provided by the Liver Cancer Institute, Fudan Ethics Committee of Zhongshan Hospital, Fudan Uni- University (Shanghai, China)40 and cultured in Dulbecco’s versity. Tissues were stored at −80 °C immediately after modified Eagle’s medium (DMEM) containing 4 mM harvesting until further processing. L-glutamine and 1 g/l glucose and supplemented with 10% FBS (Cat. # 10099, GIBCO, Carlsbad, CA, USA). A CAF and NF isolation from HCC samples and identification metastatic HCC cell line, Huh-7 (JCRB0403), was Five fresh HCC samples and their corresponding non- obtained from the Japanese Cancer Research Resources tumor controls were washed with serum-free DMEM/F- Bank (JCRB; Osaka, Japan) and cultured in DMEM con- 12 medium, cut into 0.2 × 0.2 mm fragments, and taining 4 mM L-glutamine and 1 g/L glucose and sup- incubated in a fresh culture medium for 24 h for attach- plemented with 10% FBS (Cat. # 10099, Gibco). ment. After incubation, the unattached cells were MHCC-97H cells were cultured in different conditioned removed and the remaining cells were allowed to grow on media for 72 h and evaluated at different magnifications the plate for 2–3 weeks. During this period, the medium using scanning electron microscopy (SEM; Regulus 8100, was replenished every 2 days until fibroblasts began to Hitachi, Tokyo, Japan). grow out. To ensure that there was no contamination of other cell types, we verified that the fibroblasts for >90% Cell treatment positive for the fibroblast marker α-SMA and negative for MHCC-97H and Huh-7 cells were cultured with endothelial, HCC, and epithelial cell markers, including NF-CM, CAF-CM or CAF-CM containing different

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 16 of 19

concentrations of CXCL11 recombinant protein (10876- Immunoblotting HNAE, SinoBiological, China) or CXCL11 neutralizing After 72 h of transfection, cells were added with an antibody (MAB672, R&D Systems, USA) for 72 h. Then, appropriate amount of RIPA (PMSF) buffer. Then, the total Transwell, wound healing, IF, and western blot assays protein was extracted by centrifugation, the total protein were performed. concentration was determined using the BCA method, and the protein was separated using SDS-PAGE. The proteins RNA sequencing and bioinformatics analysis were electrotransferred to PVDF membranes and blocked MHCC-97H cells were treated with or without CXCL11 with 5% milk blocking solution for 1 h to block nonspecific (10 ng/ml) for 72 h. Then, the total RNA was isolated by bindings. The membranes were incubated with the fol- TRIzol (Invitrogen). Next-generation sequencing was per- lowing primary antibodies overnight at 4 °C and washed formed by Aksomics, Inc. (Shanghai, China) using an Illu- three times with 0.05% TBST for 5 min each time: CXCL11 mina HiSeq 4000 system (USA). The differentially expressed (CSB-PA06119A0Rb; Cusabio, Wuhan, China), Vimentin genes (mRNAs and circRNAs) analysis was performed using (10366-1-AP; Proteintech, Wuhan, China), Twist (CSB- DESeq R package. The genes with threshold values of | PA025358LA01HU, Cusabio), β-actin (60008-1-Ig, Pro- log2FC| > 0.56 and a Benjamini–Hochberg corrected P value teintech), IFIT1 (CSB-PA011018LA01HU, Cusabio), and of 0.05 were selected. Finally, these differentially expressed IFIT3 (CSB-PA011022HA01HU, Cusabio). The mem- genes were subjected to Gene Ontology (GO) analysis branes were then incubated with secondary antibodies (http://geneontology.org/) and Kyoto Encyclopedia of Genes (HRP-labeled goat anti-rabbit IgG and goat anti-mouse and Genomes (KEGG) signaling annotation (https://www. IgG) at room temperature for 1 h and washed three times genome.jp/kegg/). with 0.05% TBST for 5 min each time. The signal was detected by the ECL chemiluminescence method. Cell transfection For CXCL11 knockdown in CAFs, the sh-CXCL11 H&E staining vector was obtained from GenePharma (Shanghai, China), Tissue samples were fixed in 4% paraformaldehyde, and the sh-NC vector was used as the negative control. embedded in paraffin, and cut into 4-μm thick sections. For miR-4756 overexpression or inhibition in HCC cells, H&E staining was performed to observe the histopatholo- miR-4756 mimics or inhibitor was synthesized and gical features41.Atleastfive fields were analyzed on each obtained from GenePharma; mimics NC or inhibitor NC section, and an Olympus microscope captured images. was used as a negative control. For circUBAP2, IFIT1, or IFIT3 silencing, small interfering RNA targeting circU- Immunohistochemical (IHC) staining BAP2 (si1/2-circUBAP2; si-NC as a negative control), After dewaxing and hydration, the tissue sections were IFIT1 (si1/2-IFIT1; si-NC as a negative control), or IFIT3 permeabilized with 0.2% triton (Sigma, St. Louis, MO, (si1/2-IFIT3; si-NC as a negative control) was synthesized USA) at room temperature for 10 min, and then incubated and obtained from GenePharma. Then, these vectors, with blocking solution (3.75% BSA/5% goat serum; Zymed, siRNAs, miRNA mimics, or inhibitor were transfected Carlsbad, CA, USA) for 30 min. Next, samples were incu- into target cells with Lipofectamine 3000 Reagent bated with anti-CXCL11 (CSB-PA06119A0Rb, Cusabio). (Thermo Fisher Scientific, Waltham, MA, USA) following All sections were incubated with poly-HRP antibody for the manufacturer’s protocol. The siRNA or miRNA 30 min. Then, sections were stained by DAB staining kit sequences are presented in Table S1. (Boster, Wuhan, China), and then stained with hematox- ylin (0.2 mg/ml, Sigma) to label the nucleus. Finally, sec- Real-time quantitative polymerase chain reaction (real- tions were observed under a microscope. time qPCR) Cells were digested and lysed, and total RNA was MTT assay extracted according to the RNA isolation kit instructions HCC cells (5 × 103 cells per well in 96-well plates) were (Toyobo, Tokyo, Japan). A total of 5 μl of RNA was added to each well of the 96-well plates. At each time point, obtained from each group, and ultrapure water of RNase 20 μl MTT (5 mg/ml; Sigma-Aldrich, St. Louis, MO, USA) was added to dilute the substance. Reverse transcription was then added to each well and followed by incubation for was performed according to the Fermentas RevertAid another 4 h in a humidified incubator. After removing the First Strand cDNA Synthesis kit instructions (Thermo supernatant, DMSO was used to dissolve the formazan. Scientific). Amplification reactions were performed using The OD values were determined at 450 nm. the Universal SYBR Green Master system (Roche, Basel, Switzerland). The expression levels were calculated using EdU assay −ΔΔ the 2 Ct method. All the primers used in qRT-PCR are Cells were cultured in different condition medium shown in Table S1. containing EdU solution (final concentration of 50 μM,

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 17 of 19

RioboBio, Shanghai, China) for 2 h. At the end of the into five groups: the Huh-7 cell group (n = 6), the Huh-7 incubation, cells were collected and fixed with 1 ml 4% cells + NFs group (n = 6), the Huh-7 cells CAFs (untrans- paraformaldehyde. After washing with PBS, each well was fected; n = 6) group, the Huh-7 cells + CAFs (transfected added with 1 ml 0.5% TritonX-100 and incubated at room with sh-NC; n = 6) group, and the Huh-7 cells + CAFs temperature for 10 min. After washing with PBS, each (transfected with sh-CXCL11; n = 6) group. A total of ® well was added with 1× Apollo staining agent (RioboBio) 5×105 Huh-7 cells alone or mixed with 5 × 105 NFs or 5 × and incubated at room temperature for 10 min in the 105 transfected/untransfected CAFs were suspended in dark. Then, cells were collected and resuspended with 100 μL PBS and injected into nude mouse liver. After 500 μl PBS and analyzed by flowcytometry. 2 weeks, anesthetized mice were sacrificed and the tumor volume and weight were examined. Tumor tissues were Transwell cell migration assays collected and subjected to H&E staining for the examina- The migration analysis was performed using Transwell tion of the histopathological characteristics and IHC plates without Matrigel gel in the top chambers. Target staining for CXCL11 levels in each group. cells were transfected and digested with 0.25% trypsin, For determination of the combined effects of circU- suspended in serum-free medium and counted. Then, BAP2 and miR-4756 on HCC lung metastasis, 2 × 106 200 μl(5×105 cells/ml) of transfected cells were plated in Huh-7 cells were coinfected with sh-NC/sh-circUBAP2 the upper chamber, and 600 µl of DMEM containing 10% and anti-NC/anti-miR-4756 lentivirus (GeneChem, FBS was added to the lower chamber. After 24 h of China) and injected orthotopically into the left liver lobe incubation, the upper chamber was washed twice with of mice. After 5 weeks, anesthetized mice were sacrificed, PBS. The cells on the upper surface were discarded with a in vivo tumor growth and lung metastases were detected, cotton swab and the cells left were fixed with anhydrous and the levels of related factors were examined. methanol for 30 min and stained with 0.1% crystal violet for 30 min. The number of cells passing through the ELISA basement membrane was counted under an optical The culture medium supernatant from NFs or CAFs microscope and the representative images are shown. was collected and centrifuged (1500 rpm for 5 min). Subsequently, the secretion of CXCL11 was detected Cell migration by wound healing assays using the ELISA kit (Invitrogen, USA). HCC cells were Approximately 5 × 104 target cells were plated in six-well transfected with siRNAs or miR-4756 inhibitors for 72 h. plates. Then, a pipette tip was used to draw a gap on the Then, the supernatant was collected and centrifuged cell monolayers. Mitomycin C was added to inhibit cell (1500 rpm for 5 min). Subsequently, the secretion of IL-17 proliferation and, after 24 h, cells that migrated into the and IL-1β into the conditioned medium was detected cleared section were observed under a microscope at spe- using ELISA kits following the manufacturer’s instruc- cific time points. tions (R&D Systems, Minneapolis, MN, USA).

Animal ethics Dual-luciferase reporter assay All animal experiments complied with the Guidelines IFIT1/IFIT3 3′-UTR or circUBAP2 sequence containing for the Care and Use of Experimental Animals and were the miR-4756 binding sites were fused into psiCHECK-2 approved by the Experimental Committee of Zhongshan luciferase reporter vector (Promega) to generate wild-type Hospital, Fudan University. reporter vector (wt-IFIT1/IFIT3 3′-UTR or wt-circUBAP2); moreover, site-directed mutagenesis of the miR-4756 bind- Animals ing sites in the IFIT1/IFIT3 3′-UTR or circUBAP2 sequence BALB/c nude male mice were obtained from SJA was introduced into the luciferase reporter vector to gen- Laboratory Animal Co., Ltd (Shanghai, China). Mice were erate the mutant-type reporter vector, (mut-IFIT1/IFIT3 3′- kept separately in plastic cages at 22–25 °C in a relative UTR or mut-circUBAP2). Then, 293T cells were humidity of 50–70%. Mice were housed under a constant co-transfected with miR-4756 mimics/inhibitor together light–dark cycle and were allowed free access to food with these reporter vectors using Lipofectamine 3000. The and water. luciferase activity was determined 24 h post transfection using the Dual-Luciferase Reporter System (Promega) as Establishment of an orthotopic implantation model of HCC directed by the manufacturer’sinstructions. in mice An orthotopic implantation model of HCC in BALB/c Statistics analysis mice was established according to a previous report42.In Results are representative of at least three independent general, CAFs were transfected with sh-NC or sh-CXCL11 experiments with no less than triplicates for each group vector for 72 h. Subsequently, mice were randomly divided unless speciﬁed. Two public datasets, GSE14323 and

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 18 of 19

GSE6764, were downloaded from the Gene Expression 7. Friedman,S.L.Hepaticstellatecells:protean,multifunctional,andenigmatic Omnibus and further analyzed. Data were processed with cells of the liver. Physiol. Rev. 88,125–172 (2008). 8.Friedman,S.L.Mechanismsofhepaticfibrogenesis. Gastroenterology 134, GraphPad (San Diego, California, USA) and expressed as 1655–1669 (2008). mean ± standard deviation (S.D.). A one-way analysis of 9. Lee, U. E. & Friedman, S. L. Mechanisms of hepatic fibrogenesis. Best. Pr. Res. variance (ANOVA) and Tukey’s multiple comparison test, Clin. Gastroenterol. 25,195–206 (2011). ’ t 10. Baglieri, J., Brenner, D. A. & Kisseleva, T. The role of fibrosis and liver-associated or Student s -test, were used to assess its statistical fibroblasts in the pathogenesis of hepatocellular carcinoma. Int. J. Mol. Sci. 20, significance. P value < 0.05 is considered statistically 1723 (2019). significant. 11. Lin, Z. Y. & Chuang, W. L. Hepatocellular carcinoma cells cause different responses in expressions of cancer-promoting genes in different cancer- associated fibroblasts. Kaohsiung J. Med. Sci. 29,312–318 (2013). Author contributions 12. Liu, J. et al. Cancer-associated fibroblasts promote hepatocellular carcinoma Conceptualization: S.-J.Q., Y.Y., J.F., J.Z.; funding acquisition: S.-J.Q., Y.Y.; project metastasis through chemokine-activated hedgehog and TGF-beta pathways. administration: G.L., Z.-F.Y.; writing—original draft: G.L., Z.-F.Y., J.S.; writing— Cancer Lett. 379,49–59 (2016). review and editing: S.-J.Q., Y.Y.; data curation: G.L., Z.-F.Y., S.-J.Q., Y.Y.; formal 13. Lin,Z.Y.,Chuang,Y.H.&Chuang,W.L.Cancer-associatedfibroblasts up- analysis: G.L., P.-Y.Z., C.Z.; visualization: R.-Y.G., B.-Y.S., and Z.-T.W.; all the authors regulate CCL2, CCL26, IL6 and LOXL2 genes related to promotion of cancer read, revised and approved the final manuscript. progression in hepatocellular carcinoma cells. Biomed. Pharmacother. 66, 525–529 (2012). Funding 14. Jiang, J. et al. Peri-tumor associated fibroblasts promote intrahepatic metas- This work was supported by the National Natural Science Foundation of China tasis of hepatocellular carcinoma by recruiting cancer stem cells. Cancer Lett. (NO. 81772510, NO. 82072677 and NO. 82072672); National Key Research and 404,19–28 (2017). Development Program of China (NO. 2017YFC0908101 and NO. 15. Licona-Limon,P.,Kim,L.K.,Palm,N.W.&Flavell,R.A.TH2,allergyandgroup2 2017YFC0908102); Research Programs of Science and Technology Commission innate lymphoid cells. Nat. Immunol. 14,536–542 (2013). Foundation of Shanghai (NO. 16DZ0500301 and NO.18XD1401100). 16. Akdis, M. et al. Interleukins (from IL-1 to IL-38), interferons, transforming growth factor beta, and TNF-alpha: receptors, functions, and roles in diseases. J. Allergy Data availability Clin. Immunol. 138,984–1010 (2016). Please contact the authors for data requests. 17. Lei, K. et al. The mechanism and function of circular RNAs in human diseases. Exp. Cell Res. 368,147–158 (2018). Conflict of interest 18. Zhao, X., Cai, Y. & Xu, J. Circular RNAs: biogenesis, mechanism, and function in The authors declare no competing interests. human cancers. Int. J. Mol. Sci. 20, 3926 (2019). 19. Qian, K. et al. The long non-coding RNA NEAT1 interacted with miR-101 Consent for publication modulates breast cancer growth by targeting EZH2. Arch. Biochem. Biophys. Consent for publication was obtained from the participants. 615,1–9 (2017). 20. Yu, J. et al. CircRNA-104718 acts as competing endogenous RNA and pro- Ethics approval and consent to participate motes hepatocellular carcinoma progression through microRNA-218-5p/ – All procedures performed in studies involving humans and animals were in TXNDC5 signaling pathway. Clin. Sci. 133,1487 1503 (2019). accordance with the ethical standards of Zhongshan Hospital, Fudan 21. Yang, J. et al. Vascular mimicry formationispromotedbyparacrineTGF-beta fi University, and with the 1964 Helsinki declaration. Informed consent to and SDF1 of cancer-associated broblasts and inhibited by miR-101 in – participate in the study has been obtained from participants. hepatocellular carcinoma. Cancer Lett. 383,18 27 (2016). 22.Wang,Y.G.,Liu,J.,Shi,M.&Chen,F.X.LncRNADGCR5repressesthe development of hepatocellular carcinoma by targeting the miR-346/KLF14 axis. J. Cell Physiol. 234,572–580 (2018). ’ Publisher s note 23. Torres, S. et al. Proteome profiling of cancer-associated fibroblasts identifies Springer Nature remains neutral with regard to jurisdictional claims in novel proinflammatory signatures and prognostic markers for colorectal fi published maps and institutional af liations. cancer. Clin. Cancer Res. 19,6006–6019 (2013). 24. Lau,T.S.etal.Cancercell-derivedlymphotoxin mediates reciprocal tumour- Supplementary information The online version contains supplementary stromal interactions in human ovarian cancer by inducing CXCL11 in fibro- material available at https://doi.org/10.1038/s41419-021-03545-7. blasts. J. Pathol. 232,43–56 (2014). 25. Wang, Y. et al. Cancer-associated fibroblasts promote irradiated cancer cell Received: 26 November 2020 Revised: 18 February 2021 Accepted: 19 recovery through autophagy. EBioMedicine 17,45–56 (2017). February 2021 26. Zhu,L.etal.TheeffectofC-X-Cmotif chemokine ligand 13 in cutaneous squamous cell carcinoma treated with aminolevulinic acid-photodynamic therapy. Photodiagnosis Photodyn. Ther. 26, 389–394 (2019). 27. Zhao, L. et al. Long noncoding RNA LINC00092 acts in cancer-associated fibroblasts to drive glycolysis and progression of ovarian cancer. Cancer Res. References 77,1369–1382 (2017). 1. Liu, G. et al. ROR-α-1 inhibits the proliferation, invasion, and migration of 28. Zhou, X. et al. Interferon induced IFIT family genes in host antiviral defense. Int hepatocellular carcinoma MHCC97H via downregulation of chemokine J. Biol. Sci. 9,200–208 (2013). CXCL5. Cytokine 129, 155004 (2020). 29. Fang, T. et al. Tumor-derived exosomal miR-1247-3p induces cancer- 2. Uka, K. et al. Clinical features and prognosis of patients with extrahepatic associated fibroblast activation to foster lung metastasis of liver cancer. Nat. metastases from hepatocellular carcinoma. World J. Gastroenterol. 13, Commun. 9, 191 (2018). 414–420 (2007). 30. Monnier, J. et al. CXCR7 is up-regulated in human and murine hepatocellular 3. Wang, H. & Chen, L. Tumor microenviroment and hepatocellular carcinoma carcinoma and is specifically expressed by endothelial cells. Eur. J. Cancer 48, metastasis. J. Gastroenterol. Hepatol. 28,43–48 (2013). 138–148 (2012). 4. Fattovich, G., Stroffolini, T., Zagni, I.&Donato,F.Hepatocellular carcinoma in 31. Zhang, Y. et al. CXCL11 promotes self-renewal and tumorigenicity of cirrhosis: incidence and risk factors. Gastroenterology 127,S35–S50 (2004). alpha2delta1(+) liver tumor-initiating cells through CXCR3/ERK1/2 signaling. 5. El-Serag, H. B. Hepatocellular carcinoma. N. Engl. J. Med. 365, 1118–1127 Cancer Lett. 449, 163–171 (2019). (2011). 32. Zhao, R. et al. CircUBAP2-mediated competing endogenous RNA network 6. Friedman, S. L. Molecular regulation of hepatic fibrosis, an integrated cellular modulates tumorigenesis in pancreatic adenocarcinoma. Aging. 11, response to tissue injury. J. Biol. Chem. 275, 2247–2250 (2000). 8484–8501 (2019).

Ofﬁcial journal of the Cell Death Differentiation Association Liu et al. Cell Death and Disease (2021) 12:260 Page 19 of 19

33. Wu,H.,Li,W.,Zhu,S.,Zhang,D.&Zhang,M.CircularRNAcircUBAP2regulates 38. Zhang, Z. et al. Loss of exosomal miR-320a from cancer-associated fibroblasts proliferation and invasion of osteosarcoma cells through miR-641/YAP1 axis. contributes to HCC proliferation and metastasis. Cancer Lett. 397,33–42 (2017). Cancer Cell Int. 20, 223 (2020). 39. Zhou, Y. et al. Hepatocellular carcinoma-derived exosomal miRNA-21 con- 34. Meng, L. et al. Circular RNA UBAP2 contributes to tumor growth and tributes to tumor progression by converting hepatocyte stellate cells to metastasis of cervical cancer via modulating miR-361-3p/SOX4 axis. Cancer Cell cancer-associated fibroblasts. J. Exp. Clin. Cancer Res. 37,324(2018). Int. 20, 357 (2020). 40. Tian, J. et al. New human hepatocellular carcinoma (HCC) cell line with highly 35. Zeremski, M. et al. Intrahepatic levels of CXCR3-associated chemokines cor- metastatic potential (MHCC97) and its expressions of the factors associated relate with liver inflammation and fibrosis in chronic hepatitis C. Hepatology with metastasis. Br. J. Cancer 81,814–821 (1999). 48,1440–1450 (2008). 41.Qin,D.etal.Anti-osteoporosiseffects of osteoking via reducing reactive 36. Helbig, K. J. et al. Expression of the CXCR3 ligand I-TAC by hepatocytes in oxygen species. J. Ethnopharmacol. 244, 112045 (2019). chronic hepatitis C and its correlation with hepatic inflammation. Hepatology 42. Sun, L. et al. Resolvin D1 prevents epithelial-mesenchymal transition and 39,1220–1229 (2004). reduces the stemness features of hepatocellular carcinoma by inhibiting 37. Pidugu,V.K.,Pidugu,H.B.,Wu,M.M.,Liu,C.J.&Lee,T.C.Emergingfunctions paracrine of cancer-associated fibroblast-derived COMP. J. Exp. Clin. Cancer Res. of human IFIT proteins in cancer. Front Mol. Biosci. 6, 148 (2019). 38,1–17 (2019).

Ofﬁcial journal of the Cell Death Differentiation Association